Experience and Optimization of MongoDB for Mini‑Game Operations and Cloud Integration

This article records a talk delivered at a YunJia community offline salon. The speaker, Li Xiaohui, a product lead for Tencent Cloud MongoDB, shares two main topics: (1) operational experience and optimization techniques for popular mini‑games, and (2) a cloud‑based solution that enables developers of mini‑programs and mini‑games to use Tencent Cloud services without managing underlying infrastructure.

In the first part, the speaker describes how their team has supported many hit mini‑games in production. They highlight MongoDB’s suitability for game development because it avoids schema changes, provides built‑in geospatial indexes, supports horizontal sharding for massive data, and offers native MapReduce for data‑driven operations. Specific performance optimizations are discussed, such as the one‑thread‑per‑connection model, the importance of the minConnection parameter (calculated based on TPS and shard count), and the refreshRequirement setting to avoid frequent thread switching.

A major issue addressed is slow reads from secondary replicas caused by the WiredTiger engine’s global lock during synchronization. The team proposes a patented snapshot‑based read solution and a read‑only instance architecture that dramatically reduces read latency (from ~85 ms to ~10 ms under heavy write load).

The speaker also explains the benefits of sharding: starting with a small shard and expanding horizontally as traffic grows, without requiring application changes. Additionally, they introduce table‑level rollback capabilities, which allow operators to restore individual tables without a full instance rollback—an essential feature for game incident recovery.

The second part presents a turnkey solution for mini‑program and mini‑game developers. By clicking a button in the development IDE, developers can provision a cloud stack that includes MongoDB, serverless functions, file storage, and future services such as logging and triggers. The solution abstracts away server and database management, supports multiple isolated environments (dev, test, prod), and offers features like collection import/export, flexible indexing, and fine‑grained permission control.

During the Q&A session, participants asked about monitoring, alerting, and security. The speaker confirmed that monitoring data is integrated with Elasticsearch, automatic alert policies are provided for key metrics, and security is ensured through VPC isolation, security groups, and optional at‑rest encryption for the cloud‑hosted MongoDB service.